From owner-freebsd-bugs Sun May 3 16:38:30 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id QAA24021 for freebsd-bugs-outgoing; Sun, 3 May 1998 16:38:30 -0700 (PDT) (envelope-from owner-freebsd-bugs@FreeBSD.ORG) Received: from proxy4.ba.best.com (root@proxy4.ba.best.com [206.184.139.15]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id QAA24010 for ; Sun, 3 May 1998 16:38:27 -0700 (PDT) (envelope-from dillon@backplane.com) Received: from apollo.backplane.com (apollo.backplane.com [207.33.240.2]) by proxy4.ba.best.com (8.8.8/8.8.BEST) with ESMTP id QAA07488 for ; Sun, 3 May 1998 16:04:06 -0700 (PDT) Received: (dillon@localhost) by apollo.backplane.com (8.8.8/8.6.5) id QAA04963; Sun, 3 May 1998 16:01:02 -0700 (PDT) Date: Sun, 3 May 1998 16:01:02 -0700 (PDT) From: Matthew Dillon Message-Id: <199805032301.QAA04963@apollo.backplane.com> To: Poul-Henning Kamp Cc: Nate Williams , freebsd-bugs@FreeBSD.ORG Subject: Re: i386/5398 Sender: owner-freebsd-bugs@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org Well, yes, but if it's a software blit (the only thing that would actually monopolize the PCI bus that way), it's just a tight code loop and the worst case interrupt latency is only that which is required to clear out the pci write pipeline, which is usually less then a microsecond. If it's a hardware blit, such as you get when you shift windows around, there's normally no PCI activity whatsoever and this is also true of most pattern blits since the patterns are stored in video memory. And, in anycase, the chipsets limit the burst size the dma master is allowed to do. This sort of thing cannot account for enough serial latency on its own to overrun the fifo that badly unless it's being helped a bit by interrupt disablement inside the kernel somewhere. And most blits aren't 3 or 4 MB of data anyway unless you are redrawing the whole screen. -Matt : :By monopolizing the PCI bus, while blitting 3 or 4 MB data, I can :certainly see PCI starvation. Remeber that to blt 3 MB around, you :have to read 3 MB and write 3MB, so that is 6MB over a 133MB/sec bus, :giving a minimum duration of 45 msec per blt. : :Why do you think BLT is being put into the graphics chip by so many :people ? : :Now, can we stop this futile waste of everybodys time ? : :Thankyou! : :-- :Poul-Henning Kamp FreeBSD coreteam member :phk@FreeBSD.ORG "Real hackers run -current on their laptop." :"ttyv0" -- What UNIX calls a $20K state-of-the-art, 3D, hi-res color terminal : Matthew Dillon Engineering, BEST Internet Communications, Inc. [always include a portion of the original email in any response!] To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-bugs" in the body of the message